214 million proteins compared by their shape: AI unearths TM184C and an unknown cellular communication layer

Researchers at the Sylvester Comprehensive Cancer Center sorted out predicted structures rather than genetic sequences, and uncovered a protein that was invisible to conventional methods.

Searching for a protein through its genetic sequence leaves in the shadow those whose code has diverged too much. The team at the Sylvester Comprehensive Cancer Center, at the University of Miami, reversed the method: to compare the predicted three-dimensional forms of more than 214 million proteins, and to let structural pairing report relatives that genetics no longer sees.

The approach, published in Nature, brought out TM184C, a protein related to receptors coupled with G proteins. His study revealed a previously undetected layer of cell-to-cell communication based on resource exchange via intercellular bridges, as well as a role in autophagy regulation. The function is preserved from yeast to humans, about a billion years of evolution.

The authors present less an isolated result than a method: there would be "another layer of biology that remains largely invisible", which the large-scale structural comparison makes for the first time systematically exploreable.